Set replicas to panic on disk errors, and optionally panic on replication errors (#10504)

* Till now, replicas that were unable to persist, would still execute the commands
  they got from the master, now they'll panic by default, and we add a new
  `replica-ignore-disk-errors` config to change that.
* Till now, when a command failed on a replica or AOF-loading, it only logged a
  warning and a stat, we add a new `propagation-error-behavior` config to allow
  panicking in that state (may become the default one day)

Note that commands that fail on the replica can either indicate a bug that could
cause data inconsistency between the replica and the master, or they could be
in some cases (specifically in previous versions), a result of a command (e.g. EVAL)
that failed on the master, but still had to be propagated to fail on the replica as well.

1 files changed, 25 insertions, 0 deletions

diff --git a/redis.conf b/redis.conf
index d2074f13f..62062b397 100644
--- a/redis.conf
+++ b/redis.conf
@@ -728,6 +728,31 @@ repl-disable-tcp-nodelay no
 # By default the priority is 100.
 replica-priority 100
 
+# The propagation error behavior controls how Redis will behave when it is
+# unable to handle a command being processed in the replication stream from a master
+# or processed while reading from an AOF file. Errors that occur during propagation
+# are unexpected, and can cause data inconsistency. However, there are edge cases
+# in earlier versions of Redis where it was possible for the server to replicate or persist
+# commands that would fail on future versions. For this reason the default behavior
+# is to ignore such errors and continue processing commands.
+#
+# If an application wants to ensure there is no data divergence, this configuration
+# should be set to 'panic' instead. The value can also be set to 'panic-on-replicas'
+# to only panic when a replica encounters an error on the replication stream. One of
+# these two panic values will become the default value in the future once there are
+# sufficient safety mechanisms in place to prevent false positive crashes.
+#
+# propagation-error-behavior ignore
+
+# Replica ignore disk write errors controls the behavior of a replica when it is
+# unable to persist a write command received from its master to disk. By default,
+# this configuration is set to 'no' and will crash the replica in this condition.
+# It is not recommended to change this default, however in order to be compatible
+# with older versions of Redis this config can be toggled to 'yes' which will just
+# log a warning and execute the write command it got from the master.
+#
+# replica-ignore-disk-write-errors no
+
 # -----------------------------------------------------------------------------
 # By default, Redis Sentinel includes all replicas in its reports. A replica
 # can be excluded from Redis Sentinel's announcements. An unannounced replica